summaryrefslogtreecommitdiffstats
path: root/drivers/cpuidle/governors/menu.c
blob: 03d38c291de62d7f6503f13855142fb5948b64a0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
/*
 * menu.c - the menu idle governor
 *
 * Copyright (C) 2006-2007 Adam Belay <abelay@novell.com>
 * Copyright (C) 2009 Intel Corporation
 * Author:
 *        Arjan van de Ven <arjan@linux.intel.com>
 *
 * This code is licenced under the GPL version 2 as described
 * in the COPYING file that acompanies the Linux Kernel.
 */

#include <linux/kernel.h>
#include <linux/cpuidle.h>
#include <linux/pm_qos.h>
#include <linux/time.h>
#include <linux/ktime.h>
#include <linux/hrtimer.h>
#include <linux/tick.h>
#include <linux/sched.h>
#include <linux/math64.h>
#include <linux/module.h>

/*
 * Please note when changing the tuning values:
 * If (MAX_INTERESTING-1) * RESOLUTION > UINT_MAX, the result of
 * a scaling operation multiplication may overflow on 32 bit platforms.
 * In that case, #define RESOLUTION as ULL to get 64 bit result:
 * #define RESOLUTION 1024ULL
 *
 * The default values do not overflow.
 */
#define BUCKETS 12
#define INTERVAL_SHIFT 3
#define INTERVALS (1UL << INTERVAL_SHIFT)
#define RESOLUTION 1024
#define DECAY 8
#define MAX_INTERESTING 50000


/*
 * Concepts and ideas behind the menu governor
 *
 * For the menu governor, there are 3 decision factors for picking a C
 * state:
 * 1) Energy break even point
 * 2) Performance impact
 * 3) Latency tolerance (from pmqos infrastructure)
 * These these three factors are treated independently.
 *
 * Energy break even point
 * -----------------------
 * C state entry and exit have an energy cost, and a certain amount of time in
 * the  C state is required to actually break even on this cost. CPUIDLE
 * provides us this duration in the "target_residency" field. So all that we
 * need is a good prediction of how long we'll be idle. Like the traditional
 * menu governor, we start with the actual known "next timer event" time.
 *
 * Since there are other source of wakeups (interrupts for example) than
 * the next timer event, this estimation is rather optimistic. To get a
 * more realistic estimate, a correction factor is applied to the estimate,
 * that is based on historic behavior. For example, if in the past the actual
 * duration always was 50% of the next timer tick, the correction factor will
 * be 0.5.
 *
 * menu uses a running average for this correction factor, however it uses a
 * set of factors, not just a single factor. This stems from the realization
 * that the ratio is dependent on the order of magnitude of the expected
 * duration; if we expect 500 milliseconds of idle time the likelihood of
 * getting an interrupt very early is much higher than if we expect 50 micro
 * seconds of idle time. A second independent factor that has big impact on
 * the actual factor is if there is (disk) IO outstanding or not.
 * (as a special twist, we consider every sleep longer than 50 milliseconds
 * as perfect; there are no power gains for sleeping longer than this)
 *
 * For these two reasons we keep an array of 12 independent factors, that gets
 * indexed based on the magnitude of the expected duration as well as the
 * "is IO outstanding" property.
 *
 * Repeatable-interval-detector
 * ----------------------------
 * There are some cases where "next timer" is a completely unusable predictor:
 * Those cases where the interval is fixed, for example due to hardware
 * interrupt mitigation, but also due to fixed transfer rate devices such as
 * mice.
 * For this, we use a different predictor: We track the duration of the last 8
 * intervals and if the stand deviation of these 8 intervals is below a
 * threshold value, we use the average of these intervals as prediction.
 *
 * Limiting Performance Impact
 * ---------------------------
 * C states, especially those with large exit latencies, can have a real
 * noticeable impact on workloads, which is not acceptable for most sysadmins,
 * and in addition, less performance has a power price of its own.
 *
 * As a general rule of thumb, menu assumes that the following heuristic
 * holds:
 *     The busier the system, the less impact of C states is acceptable
 *
 * This rule-of-thumb is implemented using a performance-multiplier:
 * If the exit latency times the performance multiplier is longer than
 * the predicted duration, the C state is not considered a candidate
 * for selection due to a too high performance impact. So the higher
 * this multiplier is, the longer we need to be idle to pick a deep C
 * state, and thus the less likely a busy CPU will hit such a deep
 * C state.
 *
 * Two factors are used in determing this multiplier:
 * a value of 10 is added for each point of "per cpu load average" we have.
 * a value of 5 points is added for each process that is waiting for
 * IO on this CPU.
 * (these values are experimentally determined)
 *
 * The load average factor gives a longer term (few seconds) input to the
 * decision, while the iowait value gives a cpu local instantanious input.
 * The iowait factor may look low, but realize that this is also already
 * represented in the system load average.
 *
 */

struct menu_device {
	int		last_state_idx;
	int             needs_update;

	unsigned int	next_timer_us;
	unsigned int	predicted_us;
	unsigned int	bucket;
	unsigned int	correction_factor[BUCKETS];
	unsigned int	intervals[INTERVALS];
	int		interval_ptr;
};


#define LOAD_INT(x) ((x) >> FSHIFT)
#define LOAD_FRAC(x) LOAD_INT(((x) & (FIXED_1-1)) * 100)

static inline int get_loadavg(unsigned long load)
{
	return LOAD_INT(load) * 10 + LOAD_FRAC(load) / 10;
}

static inline int which_bucket(unsigned int duration, unsigned long nr_iowaiters)
{
	int bucket = 0;

	/*
	 * We keep two groups of stats; one with no
	 * IO pending, one without.
	 * This allows us to calculate
	 * E(duration)|iowait
	 */
	if (nr_iowaiters)
		bucket = BUCKETS/2;

	if (duration < 10)
		return bucket;
	if (duration < 100)
		return bucket + 1;
	if (duration < 1000)
		return bucket + 2;
	if (duration < 10000)
		return bucket + 3;
	if (duration < 100000)
		return bucket + 4;
	return bucket + 5;
}

/*
 * Return a multiplier for the exit latency that is intended
 * to take performance requirements into account.
 * The more performance critical we estimate the system
 * to be, the higher this multiplier, and thus the higher
 * the barrier to go to an expensive C state.
 */
static inline int performance_multiplier(unsigned long nr_iowaiters, unsigned long load)
{
	int mult = 1;

	/* for higher loadavg, we are more reluctant */

	mult += 2 * get_loadavg(load);

	/* for IO wait tasks (per cpu!) we add 5x each */
	mult += 10 * nr_iowaiters;

	return mult;
}

static DEFINE_PER_CPU(struct menu_device, menu_devices);

static void menu_update(struct cpuidle_driver *drv, struct cpuidle_device *dev);

/*
 * Try detecting repeating patterns by keeping track of the last 8
 * intervals, and checking if the standard deviation of that set
 * of points is below a threshold. If it is... then use the
 * average of these 8 points as the estimated value.
 */
static unsigned int get_typical_interval(struct menu_device *data)
{
	int i, divisor;
	unsigned int max, thresh, avg;
	uint64_t sum, variance;

	thresh = UINT_MAX; /* Discard outliers above this value */

again:

	/* First calculate the average of past intervals */
	max = 0;
	sum = 0;
	divisor = 0;
	for (i = 0; i < INTERVALS; i++) {
		unsigned int value = data->intervals[i];
		if (value <= thresh) {
			sum += value;
			divisor++;
			if (value > max)
				max = value;
		}
	}
	if (divisor == INTERVALS)
		avg = sum >> INTERVAL_SHIFT;
	else
		avg = div_u64(sum, divisor);

	/* Then try to determine variance */
	variance = 0;
	for (i = 0; i < INTERVALS; i++) {
		unsigned int value = data->intervals[i];
		if (value <= thresh) {
			int64_t diff = (int64_t)value - avg;
			variance += diff * diff;
		}
	}
	if (divisor == INTERVALS)
		variance >>= INTERVAL_SHIFT;
	else
		do_div(variance, divisor);

	/*
	 * The typical interval is obtained when standard deviation is
	 * small (stddev <= 20 us, variance <= 400 us^2) or standard
	 * deviation is small compared to the average interval (avg >
	 * 6*stddev, avg^2 > 36*variance). The average is smaller than
	 * UINT_MAX aka U32_MAX, so computing its square does not
	 * overflow a u64. We simply reject this candidate average if
	 * the standard deviation is greater than 715 s (which is
	 * rather unlikely).
	 *
	 * Use this result only if there is no timer to wake us up sooner.
	 */
	if (likely(variance <= U64_MAX/36)) {
		if ((((u64)avg*avg > variance*36) && (divisor * 4 >= INTERVALS * 3))
							|| variance <= 400) {
			return avg;
		}
	}

	/*
	 * If we have outliers to the upside in our distribution, discard
	 * those by setting the threshold to exclude these outliers, then
	 * calculate the average and standard deviation again. Once we get
	 * down to the bottom 3/4 of our samples, stop excluding samples.
	 *
	 * This can deal with workloads that have long pauses interspersed
	 * with sporadic activity with a bunch of short pauses.
	 */
	if ((divisor * 4) <= INTERVALS * 3)
		return UINT_MAX;

	thresh = max - 1;
	goto again;
}

/**
 * menu_select - selects the next idle state to enter
 * @drv: cpuidle driver containing state data
 * @dev: the CPU
 */
static int menu_select(struct cpuidle_driver *drv, struct cpuidle_device *dev)
{
	struct menu_device *data = this_cpu_ptr(&menu_devices);
	int latency_req = pm_qos_request(PM_QOS_CPU_DMA_LATENCY);
	int i;
	unsigned int interactivity_req;
	unsigned int expected_interval;
	unsigned long nr_iowaiters, cpu_load;

	if (data->needs_update) {
		menu_update(drv, dev);
		data->needs_update = 0;
	}

	/* Special case when user has set very strict latency requirement */
	if (unlikely(latency_req == 0))
		return 0;

	/* determine the expected residency time, round up */
	data->next_timer_us = ktime_to_us(tick_nohz_get_sleep_length());

	get_iowait_load(&nr_iowaiters, &cpu_load);
	data->bucket = which_bucket(data->next_timer_us, nr_iowaiters);

	/*
	 * Force the result of multiplication to be 64 bits even if both
	 * operands are 32 bits.
	 * Make sure to round up for half microseconds.
	 */
	data->predicted_us = DIV_ROUND_CLOSEST_ULL((uint64_t)data->next_timer_us *
					 data->correction_factor[data->bucket],
					 RESOLUTION * DECAY);

	expected_interval = get_typical_interval(data);
	expected_interval = min(expected_interval, data->next_timer_us);

	if (CPUIDLE_DRIVER_STATE_START > 0) {
		struct cpuidle_state *s = &drv->states[CPUIDLE_DRIVER_STATE_START];
		unsigned int polling_threshold;

		/*
		 * We want to default to C1 (hlt), not to busy polling
		 * unless the timer is happening really really soon, or
		 * C1's exit latency exceeds the user configured limit.
		 */
		polling_threshold = max_t(unsigned int, 20, s->target_residency);
		if (data->next_timer_us > polling_threshold &&
		    latency_req > s->exit_latency && !s->disabled &&
		    !dev->states_usage[CPUIDLE_DRIVER_STATE_START].disable)
			data->last_state_idx = CPUIDLE_DRIVER_STATE_START;
		else
			data->last_state_idx = CPUIDLE_DRIVER_STATE_START - 1;
	} else {
		data->last_state_idx = CPUIDLE_DRIVER_STATE_START;
	}

	/*
	 * Use the lowest expected idle interval to pick the idle state.
	 */
	data->predicted_us = min(data->predicted_us, expected_interval);

	/*
	 * Use the performance multiplier and the user-configurable
	 * latency_req to determine the maximum exit latency.
	 */
	interactivity_req = data->predicted_us / performance_multiplier(nr_iowaiters, cpu_load);
	if (latency_req > interactivity_req)
		latency_req = interactivity_req;

	/*
	 * Find the idle state with the lowest power while satisfying
	 * our constraints.
	 */
	for (i = data->last_state_idx + 1; i < drv->state_count; i++) {
		struct cpuidle_state *s = &drv->states[i];
		struct cpuidle_state_usage *su = &dev->states_usage[i];

		if (s->disabled || su->disable)
			continue;
		if (s->target_residency > data->predicted_us)
			continue;
		if (s->exit_latency > latency_req)
			continue;

		data->last_state_idx = i;
	}

	return data->last_state_idx;
}

/**
 * menu_reflect - records that data structures need update
 * @dev: the CPU
 * @index: the index of actual entered state
 *
 * NOTE: it's important to be fast here because this operation will add to
 *       the overall exit latency.
 */
static void menu_reflect(struct cpuidle_device *dev, int index)
{
	struct menu_device *data = this_cpu_ptr(&menu_devices);

	data->last_state_idx = index;
	data->needs_update = 1;
}

/**
 * menu_update - attempts to guess what happened after entry
 * @drv: cpuidle driver containing state data
 * @dev: the CPU
 */
static void menu_update(struct cpuidle_driver *drv, struct cpuidle_device *dev)
{
	struct menu_device *data = this_cpu_ptr(&menu_devices);
	int last_idx = data->last_state_idx;
	struct cpuidle_state *target = &drv->states[last_idx];
	unsigned int measured_us;
	unsigned int new_factor;

	/*
	 * Try to figure out how much time passed between entry to low
	 * power state and occurrence of the wakeup event.
	 *
	 * If the entered idle state didn't support residency measurements,
	 * we use them anyway if they are short, and if long,
	 * truncate to the whole expected time.
	 *
	 * Any measured amount of time will include the exit latency.
	 * Since we are interested in when the wakeup begun, not when it
	 * was completed, we must subtract the exit latency. However, if
	 * the measured amount of time is less than the exit latency,
	 * assume the state was never reached and the exit latency is 0.
	 */

	/* measured value */
	measured_us = cpuidle_get_last_residency(dev);

	/* Deduct exit latency */
	if (measured_us > 2 * target->exit_latency)
		measured_us -= target->exit_latency;
	else
		measured_us /= 2;

	/* Make sure our coefficients do not exceed unity */
	if (measured_us > data->next_timer_us)
		measured_us = data->next_timer_us;

	/* Update our correction ratio */
	new_factor = data->correction_factor[data->bucket];
	new_factor -= new_factor / DECAY;

	if (data->next_timer_us > 0 && measured_us < MAX_INTERESTING)
		new_factor += RESOLUTION * measured_us / data->next_timer_us;
	else
		/*
		 * we were idle so long that we count it as a perfect
		 * prediction
		 */
		new_factor += RESOLUTION;

	/*
	 * We don't want 0 as factor; we always want at least
	 * a tiny bit of estimated time. Fortunately, due to rounding,
	 * new_factor will stay nonzero regardless of measured_us values
	 * and the compiler can eliminate this test as long as DECAY > 1.
	 */
	if (DECAY == 1 && unlikely(new_factor == 0))
		new_factor = 1;

	data->correction_factor[data->bucket] = new_factor;

	/* update the repeating-pattern data */
	data->intervals[data->interval_ptr++] = measured_us;
	if (data->interval_ptr >= INTERVALS)
		data->interval_ptr = 0;
}

/**
 * menu_enable_device - scans a CPU's states and does setup
 * @drv: cpuidle driver
 * @dev: the CPU
 */
static int menu_enable_device(struct cpuidle_driver *drv,
				struct cpuidle_device *dev)
{
	struct menu_device *data = &per_cpu(menu_devices, dev->cpu);
	int i;

	memset(data, 0, sizeof(struct menu_device));

	/*
	 * if the correction factor is 0 (eg first time init or cpu hotplug
	 * etc), we actually want to start out with a unity factor.
	 */
	for(i = 0; i < BUCKETS; i++)
		data->correction_factor[i] = RESOLUTION * DECAY;

	return 0;
}

static struct cpuidle_governor menu_governor = {
	.name =		"menu",
	.rating =	20,
	.enable =	menu_enable_device,
	.select =	menu_select,
	.reflect =	menu_reflect,
	.owner =	THIS_MODULE,
};

/**
 * init_menu - initializes the governor
 */
static int __init init_menu(void)
{
	return cpuidle_register_governor(&menu_governor);
}

postcore_initcall(init_menu);